Historically margin levels have been static and easy to predict, however, the introduction of new regulation has led to an increase in volatility in margin levels. This volatility has come from a combination of more focus on real time risk and the introduction of new procedures and processes that the CCPs need to follow.
These changes have led to margins that can be seen to jump by as much as 40% on a daily basis (shown below).
In the wake of the 2008 financial crisis, greater scrutiny was placed on CCPs and their central role in risk management. This led to a proliferation of new regulation to which they needed to adhere. Much of this focussed on the way that margin was calculated and the risks that it needed to cover.
Historically, margin algorithms were expected to cover the market risk under normal market conditions, backed by a default fund to support this that covered losses under stress scenarios.
Now margin needs to cover a number of additional risks including liquidity risk and credit risk. And it needs to take care not to be procyclical – basically raising the margin because of increased market volatility leading to even more volatility. Not to mention the fact it needs to be recalculated intra-day rather than just being an end of day calculation and that parameters must be reviewed on a more regular basis.
How margin calculations have been impacted
Consequently, margin algorithms are now more complex due to multiple add-ons each with parameters that need to be set. Each time these change the margin calculated will change, leading to more volatile margin levels. When it was just a scenario based calculation of potential change in market levels or rates it was relatively easy to predict how margin might change given the market conditions. But now there are a number of factors to consider, including the following:
- Intra-day margin
- Regular margin reviews
- Increased use of VaR as a margin methodology.
Why each of these may increase the volatility and unpredictability of margin levels is described below:
It used to be easy to estimate margin. With an algorithm like SPAN there would be a margin rate and a margin estimate could be made by multiplying this rate by the position. The rate would change predictably, maybe once every three months. But now everything is much more complicated.
An intra-day margin call used to be a rare thing. They only happened in very volatile markets, and the additional margin requested would be a simple multiple of the EOD margin. However, now they are based on real time calculation of the risk, including both Initial Margin and estimated Variation Margin.
This need for the CCPs to be covering their risk in real-time means that any new trades or market movement could result in a request to post additional collateral. The amounts requested can be unpredictable as the CCPs will snap market prices and positions and then perform the calculation at a given point in time. And if this snap happens, say, just as a price peaks or when only half of a risk neutral strategy has been submitted for clearing, then the request can be larger than expected. Once the market has settled and the rest of the position has been submitted, the margin will drop back to the expected level. But this change will only occur at the end of day, with what could feel like an unexplained return of cash.
Here’s one example for the FTSE 100 future. The Initial Margin for this contract is calculated using SPAN so will remain constant until the margin parameters are changed.
If you have a position of long 100 lots then the initial margin will be:
Number of lots * SPAN Scanning Range = 100 * 3071 = £307,100
So, assume at the end of the previous day, the settlement price was 7,000. However, during the day the price dropped to 6,850 (a 150 point drop in the FTSE). As a long position holder you will have a theoretical loss of:
Number of lots * price move * contract size = 100 * 150 *10 = £150,000
Whereas historically, when the price move was less than 50% of the Initial Margin (current rate is a 307.1 point move) this would have been ignored, now you will be required to cover this theoretical loss intra-day.
Yet, the level of the FTSE will probably have recovered by the end of day. Therefore, let’s assume the closing settlement price is 7010. When the actual variation margin is calculated it will be:
Number of lots * price move * contract size = 100 * (7010-7000) * 10 = £10,000
That is, you have now made a profit based on the previous night’s closing price. The £150,000 you had to provide intra-day will be returned along with the £10,000 profit.
Regular Margin Reviews
Historically, unless there was a large market move, SPAN margin parameters were reviewed about once every three months. Now, for those exchanges still using SPAN, this is an almost continuous process. For a large CCP like ICE there could be changes to parameters, based on ad-hoc and regular reviews, multiple times a week. As an example, over the past month, there have been six separate updates to the margin parameters for ICE energy products.
Many CCPs are using an historic VaR based algorithm, which means that the margin is changing on a daily basis. They also need to run daily backtesting, with any breaches investigated. This can result in changes to the VaR parameters, for example adjusting the confidence level or adding new stress scenarios, again resulting in unpredictable changes in margin.
As an example, during the recent Italian crisis, margins at Eurex for the BTP (Italian bond) futures contract jumped by over 40%. This increase was caused by additional scenarios that had been added, presumably to cover observed losses. This sudden change was for long position holders only – short position holders saw only a very slight increase in margin, because they were not being impacted by the fall in Italian bond prices.
The parameters for the market risk component of the margin aren’t the only ones that can change. CCPs will also review the parameters for the other parts of the margin, for example the liquidity charge.
As an example, Eurex recently updated all of their liquidity parameters for OTC products. Many of these changes were small, but for large positions a wholesale change was made to the parameters, doubling the size of the liquidity factor, which had previously been set at the same rate as for lower volumes.
Use of VaR
In the ETD markets, the new regulation has mandated CCPs to enhance their margin algorithms in order to be compliant. In many cases this has included moving away from a SPAN style margin to using VaR as the basis of their algorithm. For example, LME Clear are currently planning to move to VaR margining in 2019.
Additionally, whereas the margin that is calculated using a scenario based algorithm like SPAN is relatively stable, historic VaR is inherently more volatile. There are a number of factors that lead to this:
Most CCPs are using Filtered Historical Simulation, which means that the VaR scenarios are adjusted to reflect current volatility, and hence will change on a daily basis.
For some products, CCPs will be using relative scenarios, which means that the profit or loss for each scenario will change as the underlying prices change.
A moving window of scenarios will be used, so over time scenarios will be dropped whilst new ones are added. And any one of these could be one that results in a loss included in the VaR calculation.